1. Field of the Invention
The present invention relates to a fingerprint authentication device and a computer readable medium, and in particular, relates to a fingerprint authentication device that carries out correction of pixel values of acquired fingerprint image data and carries out fingerprint authentication in accordance with the spectral data method by using the corrected image data and to a computer readable medium that stores a program for fingerprint authentication.
2. Description of the Related Art
Fingerprint authentication is used as one type of biometric authentication. Mainly the following three methods are known for fingerprint authentication.
The first method is pattern comparison. This method verifies a fingerprint by comparing a portion of inputted image data of a fingerprint with a template recorded in a device.
The second method is minutiae comparison that finds characteristic points from inputted image data of a fingerprint, and compares these characteristic points with minutia points of a template recorded in a device.
The third method is the spectral data method that divides an image of a fingerprint, that is inputted from a line sensor or an area sensor, into horizontally-long rectangular images, and thereafter, for each of the rectangular images, generates spectral data by digitizing information such as the directions of the mountain portions (ridges) of the fingerprint, and the frequency that expresses how many ridges there are in a predetermined range, and the like.
Further, the spectral data obtained at the respective rectangular images is plotted as a matrix so as to generate a spectral data matrix, and the obtained spectral data matrix is compared with a registered spectral data matrix for verification that is recorded in a device, and authentication of the fingerprint is carried out.
FIG. 1 is a schematic drawing of a fingerprint authentication device in accordance with the prior art that employs the spectral data method.
As shown in FIG. 1, a fingerprint authentication device in accordance with the related art that employs the spectral data method has: a fingerprint input device 101 that is a fingerprint sensor that scans or images a fingerprint and acquires fingerprint image data; a fingerprint image storage section 102 that stores this fingerprint image data; a spectral data generation section 103 that computes the directions of the mountain portions of the fingerprint and the frequency of the fingerprint in the fingerprint image data stored in the fingerprint image storage section 102, and generates a spectral data matrix; a spectral data matrix storage section 104 that stores the spectral data matrix; a registered spectral data matrix archive section 105 that archives a registered spectral data matrix that is referred to at the time of fingerprint authentication; a fingerprint verification section 106 that verifies the spectral data matrix stored in the spectral data matrix storage section 104 with the registered spectral data matrix that is archived in the registered spectral data matrix archive section 105; and an authentication results outputting section 107 that outputs results of authentication by the fingerprint verification section 106.
The fingerprint input device 101 is a line sensor or an area sensor, and scans or images the fingerprint of a finger that is pushed on a sensor portion, and acquires fingerprint image data such as shown in the lower left of FIG. 2.
The fingerprint image data acquired by the fingerprint input device 101 is stored in a fingerprint image storage section 102.
The fingerprint image data stored in the fingerprint storage section 102 is read-out by the spectral data generation section 103, and the spectral data generation section 103 generates a spectral data matrix from the read-out fingerprint image data. The generated spectral data matrix is stored in the spectral data matrix storage section 104.
The fingerprint verification section 106 compares a registered spectral data matrix, that has been previously stored in the registered spectral data matrix archive section 105, and the spectral data matrix, that has been generated by the spectral data generation section 103 and stored in the spectral data matrix storage section 104, and, from the proportion of similar spectral data, judges matching or non-matching of the spectral data matrices.
In the spectral data method, the data of the matrices used in verification uses less memory than images of fingerprints, and comparison and verification of digitized data is carried out. Therefore, there is the advantage that the algorithm can be simplified. An invention relating to a fingerprint verification method in accordance with this method and a device therefor is disclosed in Japanese Patent Application National Publication No. 2008-529156.
Further, Japanese Patent Application Laid-Open (JP-A) No. 08-272953 discloses an invention that, when a pixel value of fingerprint image data exceeds a threshold value, adjusts the density of the image by adding or subtracting that pixel value.
When acquiring an image of a fingerprint by using a line sensor or an area sensor at a fingerprint authentication device, there are cases in which a finger is pressed strongly on the sensor portion, and therefore, an image in which the ridges of the fingerprint are squashed and thickened is inputted as shown in the bottom right of FIG. 2.
In the fingerprint image data that the sensor reads, the ridges that are the mountain portions of the fingerprint are displayed by black lines, and the valley portions of the fingerprint are displayed in white. In the fingerprint image data at the bottom right of FIG. 2, the ridges are displayed thick as compared with the fingerprint image data at the lower left of FIG. 2, and there is a state in which it is difficult to differentiate between the ridges and the valley portions.
This phenomenon in which the ridges of a fingerprint are squashed and become thicker can occur, in addition to when a finger is pressed strongly on the sensor portion, also due to differences in the sensitivity of the sensor or problems with adjustment of the sensor.
In any of these cases, the thickness of the ridges of the fingerprint and the overall density of the fingerprint image data are different from fingerprint image data that is acquired by the finger not being pressed strongly on the sensor portion. In an image that is used in fingerprint authentication, at the ridges of the fingerprint and the portions other than the ridges, the pixel values should naturally fluctuate gradually from dark to light, as shown in the lower left of FIG. 2. However, in the image shown in the lower right of FIG. 2 in which the ridges of the fingerprint are squashed and thickened, in addition to the ridges being displayed as thick, the gradual fluctuations in the pixel values from dark to light are not distinct. As a result, it is easy for differences to arise between the fingerprint image data of the lower left of FIG. 2 and the fingerprint image data of the lower right of FIG. 2, especially in the frequency data thereof when these fingerprint image data are made into spectral data.
Even in the case of frequency data that is generated from fingerprint image data relating to a same fingerprint, if the numerical values of the frequencies are different, the spectral data matrices that include the frequency data also differ. In such a case, the probability that the fingerprint verification device will erroneously judge that the frequency data do not match is high, and appropriate fingerprint authentication cannot be realized.
Although an invention that adjusts the density of an image is disclosed in JP-A No. 08-272953, the adjusting of the density of an image of the invention relating to JP-A No. 08-272953 is merely the adding or subtracting of a pixel value that exceeds a threshold value. In this method, pixel values are added equally to the portions that are squashed black of the image at the lower right in FIG. 2, and these portions are corrected so as to be lighter uniformly, and there is the concern that the contrast of the image, which is needed in fingerprint authentication, will suffer.